This Is Not How You Make a Doc About AI


Artificial intelligence (AI) has rapidly emerged as one of the hottest of hot-button issues in contemporary discourse, as its potential ramifications touch on everything from business to technology and art. Anyone now can use AI-based tools to write essays or create images; granted, for now, these are poorly written essays and ugly, soulless images, but who knows how the technology may evolve. Over time, could AI grow complex enough to develop a consciousness, or allow humans to upload their consciousnesses onto computers? Such concepts are at the heart of two documentaries premiering at this year’s Sundance Film Festival. Hans Block and Moritz Riesewieck’s Eternal You looks at the emerging “digital afterlife” industry, while Peter Sillen’s Love Machina follows an entrepreneur who wants to build a virtual facsimile of her wife. But frustratingly, both are asking the wrong questions.

Love Machina is a hagiographic profile of Martine Rothblatt, who among other things co-founded Sirius XM and United Therapeutics. The lynchpin of her ventures into extending human life via technology is BINA48, a chatbot communicating through a robotic bust that’s modeled on the head of Rothblatt’s wife Bina. This is supposedly the first step to eventually perpetuating a human mind in digital form, but the evidence for this is sorely lacking. BINA48 often meets questions with non-sequiturs or loosely relevant answers. The film depicts this without one evident crumb of skepticism. Allegedly a tribute to the Rothblatts’ love story, it often feels more like a recruitment video for their “Terasem Movement,” a hazily focused futurist project that comes across as just as creepily cult-like as the name suggests.

Eternal You is more circumspect about its topic but still not nearly critical enough. It surveys several ventures that claim to be able to create AI doubles of the deceased, chiefly Project December. There’s a lot of chatter from talking heads about the possible ramifications of such technology, all ignoring the problem that the technology as it is now does not do at all what it claims to. The film is full of people having heartfelt correspondences with “loved ones” who are obviously bots speaking in genericisms. The film is missing the forest for the trees.

The lay comprehension of AI and its attendant issues is woefully deficient. The average person’s understanding of AI is shaped by pop culture, and we are besotted with ideas of friendly Star Wars robots or threatening 2001: A Space Odyssey computers. A lot of people seem to read about things like ChatGPT and think we’ve already reached such a stage of consciousness, or are close to it. It doesn’t help that many journalists are overly credulous about the claims that Silicon Valley CEOs make concerning the capabilities and potential of AI: One would think that the rapid cycles of boom and bust seen with “Web 3.0,” NFTs, and the Metaverse over just the past few years would inspire more skepticism, but apparently not. 

This is a problem that Eternal You and Love Machina both enthusiastically perpetuate. Neither takes more than the bare minimum of time to explain how their respective subjects are supposed to work. BINA48 allegedly uses the real Bina’s memories as a template for its intelligence, but the process by which her personal materials are input into it is completely obfuscated from the audience. Meanwhile, while one has to pay to generate a “personality simulation” from Project December, the application form is freely available. It’s difficult to see such a basic questionnaire and believe that it can form the basis of an entire personality, much less one that’s supposed to mimic an actual former human.  

One of the more well-known thought experiments concerning artificial intelligence is Alan Turing’s 1950 “imitation game,” in which a machine’s believability as a thinking being is tested by seeing whether it can fool a human into believing they’re talking to another human. There’s a major flaw in the Turing test, however: The presupposition that a person could only be fooled by a sophisticated program. In practice, it turns out it doesn’t take much for someone to read humanity in a computer, even if they know that they’re talking to a computer. People who interacted with the rudimentary 1960s language processing program ELIZA treated it as if it were a person, and it possessed a minuscule fraction of the power of the large language models that undergird contemporary chatbots. 

Now we see Google engineers conversing with chatbots and believing the latter have souls based on performances that aren’t even all that convincing. Again and again, people in the two films interact with BINA48 or personality simulations with a lack of skepticism that frankly shocked me. Neither film interrogates this impulse, leaving the implicit suggestion that they are empathizing with a legitimate being. My takeaway is that whether out of naivete or cynicism, the AI projects featured in these documentaries are exploiting uninformed, often vulnerable people. The filmmakers were not obligated to come to the same conclusions, of course, but their lack of pushback forecloses few perspectives other than what their subjects deliver to the audience.

The answer to “Could AI grow complex enough to develop its own consciousness, or allow humans to upload their consciousnesses into computers?” is a firm “probably not.” As Robert Epstein explains, one recurring fallacy throughout history is that we think we come up with appropriate metaphors for the workings of the brain only for those metaphors to be discarded as we refine our engineering skills. Thousands of years ago there was a “hydraulic model of human intelligence” arguing that the brain functioned via the flow of the body’s assumed different fluids. Now we “know better” and think of brains as being like computers, which is also inaccurate. Brains function nothing like programs, which are essentially complex pattern-recognition devices. How could such a framework sustain a human-like consciousness? After all, consciousness is probably not a tangible “thing,” but rather a process whose components we still only haphazardly understand. 

All speculation and debate about AI that presume a transhumanist future just on the horizon are operating from a ludicrous position that’s useless outside the realm of pure science fiction, completely disconnected from the real issues at hand. Said issues may look disappointingly mundane, but they are no less important, particularly regarding people’s jobs and the labor market. AI has so far proved adept at facilitating plagiarism and filling the internet with ugly noise; in other words, it’s furthering the ways capitalism has already degraded and devalued creative work. And that’s to say nothing of the ways in which AI is a Potemkin technology, with the actual intelligence behind these tools often being underpaid humans. Love Machina and Eternal You are doing nothing but reinforcing that edifice, distracting from the true issues at hand. Any film, article, or book about AI that does not cut through the thick reality distortion field conjured by the technology’s boosters will fall into the same trap.

Eternal You and Love Machina (both 2024) are screening in person and virtually this month as part of the Sundance Film Festival.  



Source link

About The Author

Scroll to Top